Title: EM algorithms for estimating the B-spline copula
Authors: Xiaoling Dou - Waseda University (Japan) [presenting]
Satoshi Kuriki - The Institute of Statistical Mathematics (Japan)
Abstract: The B-spline copula is a generalization of the Bernstein copula. It is defined by replacing the Bernstein basis functions by B-spline basis functions. This change requires the copula parameters satisfy slightly different conditions, in spite of the copula form remains the same. Because the Bernstein copula can be considered as a finite mixture distribution for given marginals, we can use EM algorithm methods to estimate the Bernstein copula. Since this idea is also available for the B-spline copula, we propose to generate the existing EM algorithms of the Bernstein copula to estimate the B-spline copula by changing the basis functions and the parameter conditions. Illustrative examples are presented with real data sets.